CN110367973B

CN110367973B - Method for detecting initial point of QRS wave group of multi-lead electrocardiosignal

Info

Publication number: CN110367973B
Application number: CN201910584153.8A
Authority: CN
Inventors: 杨翠微; 何凯悦; 陈家曦; 丁小曼
Original assignee: Fudan University
Current assignee: Fudan University
Priority date: 2019-07-01
Filing date: 2019-07-01
Publication date: 2021-10-26
Anticipated expiration: 2039-07-01
Also published as: CN110367973A

Abstract

The invention relates to a method for detecting a QRS wave group starting point of a multi-lead electrocardiosignal. Collecting multi-lead electrocardiosignals by using an electrocardio collection technology; preprocessing the multi-lead electrocardiosignals to obtain signals with baseline drift and high-frequency noise removed; normalizing the signal to obtain a normalized multi-lead electrocardiosignal; calculating the root mean square RMS and a detection function DEF of the normalized multi-lead electrocardiosignals, calculating a threshold value by learning values of the RMS and the DEF several seconds before, and taking a DEF peak value point meeting the threshold value condition as the starting point of a QRS wave group. The method has the advantages of high detection accuracy, simple algorithm, real-time operation and the like. The method is suitable for long-range or short-range electrocardiosignals, can be used for analyzing sinus rhythm signals or arrhythmia signals, and has certain application value in electrophysiological mechanism research and clinical medicine. The method can be popularized to the activation analysis and related quantitative research of the intracardiac electric signals.

Description

Method for detecting initial point of QRS wave group of multi-lead electrocardiosignal

Technical Field

The invention relates to a method for detecting a QRS wave group starting point of a multi-lead electrocardiosignal.

Background

In the electrocardiographic periodic signal, the amplitude of the QRS complex is the largest and the characteristics are the most obvious, so in the existing electrocardiographic signal characteristic point detection algorithm, there are many detection methods related to the QRS complex, and the common methods include: template matching, difference threshold, artificial neural network, and wavelet transform.

The template matching method needs to make corresponding adjustment according to the difference of different individuals, the algorithm has strong dependence, and is easy to be interfered by noise, and the detection result is often not ideal. When arrhythmia occurs, particularly ventricular premature beat, the amplitude of R wave of QRS wave group is greatly changed, RR interval is irregular, and the traditional differential threshold method cannot accurately locate. Although the method of the artificial neural network has high adaptivity and a good detection effect, the realization of the method needs a large number of representative samples and is difficult to realize in the practical application process. The wavelet transform method has large calculation amount, complex steps and long time consumption, and cannot position a start point and a stop point.

The above QRS complex detection method is mainly used to extract R waves, but detection of characteristic points such as start points and end points of QRS complexes is simple, and is mainly obtained by searching forward or backward based on the peak point of R waves. In fact, the starting point and the ending point of the QRS complex are key points for correctly characterizing the electrical activity of the heart, and are of great significance for clinically diagnosing cardiogenic abnormalities (such as shock, angina pectoris, ventricular block, and the like).

Disclosure of Invention

Aiming at the characteristics of poor real-time performance, complex algorithm, low accuracy and the like of the traditional characteristic point detection algorithm, the invention aims to provide the method for detecting the initial point of the QRS complex of the multi-lead electrocardiosignal.

The invention provides a method for detecting a QRS wave group starting point of a multi-lead electrocardiosignal, which comprises the following specific steps of:

(1) collecting multi-lead electrocardiosignals by using an electrocardio collection technology;

(2) preprocessing the multi-lead electrocardiosignals obtained in the step (1) to obtain multi-lead electrocardiosignals with baseline wander and high-frequency noise removed;

(3) carrying out normalization processing on the preprocessed multi-lead electrocardiosignals obtained in the step (2) to obtain multi-lead normalized electrocardiosignals;

(4) calculating the root mean square RMS of the multi-lead normalized electrocardiosignals obtained in the step (3) to obtain RMS signals, and assuming that n-lead electrocardiosignals exist (n is more than or equal to 2), calculating the root mean square as shown in a formula (1):

wherein: ECG (ECG)_iThe amplitude of the ith lead electrocardiosignal is more than or equal to 1 and less than or equal to n;

(5) calculating a detection function DEF (detection function) by using the RMS signal obtained in the step (4), wherein the DEF is calculated as shown in the formula (2):

for the RMS signal of the electrocardiosignal obtained in the step (4), setting an o point as a starting point of a QRS wave to be detected, setting a width as a time parameter (empirical value), and setting k and m as normal numbers (empirical values); wherein:

a is the amplitude difference between the lowest point and the point o in the forward width time of the point o, if the point o is an inflection point, the value of a is close to 0, and exp (-ka) is close to 1; if point o is a certain point of the ascending branch, a is a positive value, and the value of exp (-ka) is close to 0, which is beneficial to distinguish point o from other points at the ascending branch of the RMS signal;

b is the amplitude difference between the highest point and the o point appearing in the backward width time of the o point, mainly reflects the rising speed of the RMS signal, is favorable for distinguishing the o point from a flat section in front of the o point and is also favorable for distinguishing the o point from the starting points of P waves and T waves;

c is the maximum value of the RMS signal which appears m width time after the o point, which is beneficial to distinguishing QRS wave, P wave and T wave;

d is the time difference between o and the RMS maximum occurring m width times back, generally speaking, ventricular premature beats with a larger c value but a smaller d value; the c value of sinus rhythm is small, and the d value is large; therefore, the comprehensive consideration of c and d is favorable for reducing the difference of DEF detection function values of different beats, and meanwhile, the value of d is also favorable for distinguishing QRS waves from partial slowly rising T waves;

e is the RMS integral value m width time from point o to point o;

(6) calculating the peak value of the root mean square signal in the first seconds by using the data of the first seconds of the root mean square signal obtained in the step (4) to form a group of sequences, and taking the numerical value corresponding to a certain percentage of the median value of the sequences as an RMS threshold value; calculating the peak value of the detection function signal within the first seconds by using the data of the first seconds of the detection function signal obtained in the step (5) to form a group of sequences, and taking the numerical value corresponding to a certain percentage of the median value of the sequences as a DEF threshold value; taking DEF peak points which meet a threshold condition and are not in a preset refractory period (adjustable in 200-400 ms) as QRS starting points; if QRS starting point can not be detected within 2s, the threshold value condition is relaxed, the detection is carried out again, and DEF peak point of a numerical value corresponding to a certain percentage higher than the threshold value is regarded as the starting point of QRS.

The invention has the following beneficial effects:

1. compared with the traditional detection method, the method for detecting the QRS wave group starting point of the multi-lead electrocardiosignal has the advantages of high detection accuracy rate, strong detection real-time property, simple algorithm and the like.

2. The method has certain adaptability to the detection of the starting point of the QRS wave group of the multi-lead electrocardiosignal of the abnormal heart rhythm with polymorphic ventricular premature beat, atrial fibrillation, ST segment change, bigeminy or other irregular rhythms.

3. The method is suitable for short-range or long-range electrocardiosignals, sinus rhythm signals and arrhythmia signals, and provides an important detection method for electrocardio diagnosis.

4. After the initial point is detected, the multi-lead electrocardiosignals are respectively integrated, and the obtained result can represent an average projection vector of the electrocardio comprehensive vector in each lead in a cardiac cycle, thereby being beneficial to researching arrhythmia signals such as ventricular premature beat and the like.

5. The method can be popularized to the excitation analysis and related quantitative research of the intracardiac electric signals.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments will be briefly described below. It is noted that the following drawings illustrate only certain embodiments of the invention and are therefore not to be considered limiting of its scope.

FIG. 1 is a diagram of the root mean square signal of a body surface 12 lead electrocardiosignal. Wherein, the o point is the starting point of the QRS wave to be detected; width is a set time parameter; a is the amplitude difference between the lowest point and the point o appearing in the forward width time of the point o; b is the amplitude difference between the highest point and the o point occurring in the backward width time of the o point; c is the maximum value of the RMS signal occurring 4 width times after point o; d is the time difference between o and the RMS maximum occurring 4 × width times back; e is the root mean square integral over a time period of 4 width after point o.

Fig. 2 is a schematic diagram of detection of an initial point of a QRS complex of a body surface 12-lead electrocardiosignal of a sinus rhythm. Wherein, (a) is a section of 12-lead original electrocardiosignals with length of 10s, I, II and III are standard limb leads, aVR, aVL and aVF are compression limb leads, and V1-V6 are 6 chest leads; (b) filtering the 12-lead electrocardiosignal of the segment, wherein the marks in the figure are the same as (a); (c) the normalized signal of the 12-lead electrocardiosignal of the section is represented by the ordinate of the normalized signal and the abscissa of the normalized signal is time; (d) the root mean square value of the 12-lead normalized signal is shown, the ordinate is the amplitude of the normalized signal, the abscissa is time, and the horizontal dotted line represents the root mean square threshold; (e) the detection function value of the 12-lead electrocardiosignal of the section is shown, the ordinate is the amplitude of the electrocardiosignal, the abscissa is time, and the horizontal dotted line represents the threshold value of the detection function; (f) for the final detection effect, the first 12 rows of signals are shown in (b), the 13 th row of signals are shown in (d), the 14 th row of signals are shown in (e), the vertical dotted line is the marked R wave, the vertical solid line is the starting point of the detected 12-lead electrocardiographic signal mean QRS complex, and the number at the top end of each vertical solid line represents the heart beat number.

Fig. 3 is a schematic diagram of the detection of the initial point of the QRS complex of the superficial 12-lead electrocardiosignal of the polymorphic ventricular premature beat. The picture layouts of (a) to (f) are the same as those of (a) to (f) shown in fig. 2.

Fig. 4 is a schematic diagram of detection of an initial point of a QRS complex of a body surface 12-lead electrocardiosignal of atrial fibrillation. The picture layouts of (a) to (f) are the same as those of (a) to (f) shown in fig. 2.

Fig. 5 is a schematic diagram of the detection result of the initial point of the QRS complex of the body surface 12 lead electrocardiographic signal with depressed ST segment. The picture layout is the same as fig. 2 (f).

Fig. 6 is a schematic diagram of the detection result of the initial point of the QRS complex of the body surface 12-lead electrocardiograph signals of paroxysmal ventricular rate. The picture layout is the same as fig. 2 (f).

Fig. 7 is a schematic diagram of the detection result of the initial point of the QRS complex of the superficial 12-lead electrocardiosignal of the ventricular premature beat bigeminy. The picture layout is the same as fig. 2 (f).

Detailed Description

The method and application of the present invention will be further explained based on the conventional body surface 12 lead electrocardiographic data (i.e. lead number n is 12) with reference to the drawings and the embodiments.

Example 1: the method for detecting the QRS wave group starting point is applied to sinus rhythm signals (different parts have consistent rhythms). In this embodiment, a body surface 12 lead electrocardiographic signal with a sampling rate of 1000Hz (i.e. the lead number n is 12) is used, and the working flow is as follows:

(1) the body surface 12-lead cardiac signal (digital signal) of the sinus rhythm synchronously acquired is observed, as shown in fig. 2 (a). The light grey horizontal line of each lead represents its zero potential line. It can be seen from the figure that there are some leads with serious baseline wander, such as I, II, III, aVR, aVL, aVF, V2, V3, V4, V5, and V6.

(2) And (3) preprocessing the body surface 12-lead electrocardiosignals of the sinus rhythm in the step (1). Firstly, performing 10-layer decomposition on signals by using sym4 wavelet basis functions, regarding the electrocardiosignals with 1000Hz, and setting the 10 th-layer decomposition signals (the frequency range is 0-0.98 Hz) as a base line to be 0; then, the electrocardiosignals (125-500 Hz) of the 1 st-2 nd layers are subjected to hard threshold processing to remove high-frequency noise. The body surface 12 lead electrocardiographic signals obtained by the above-mentioned pretreatment are shown in fig. 2 (b).

(3) In order to ensure that each lead contributes the same to the final detection result, the preprocessed 12-lead electrocardiograph signals obtained in step (2) are normalized to obtain signals shown in fig. 2 (c).

(4) And (4) calculating the root mean square of the normalized 12-lead electrocardiosignals obtained in the step (3) to obtain signals shown in fig. 2 (d). The peak value of the root mean square signal in the first 10s is obtained to form a sequence, and one half of the median value of the sequence is taken as the root mean square threshold value, as shown by the horizontal dotted line in the figure.

(5) The detection function of the root mean square signal obtained in step (4) is calculated (let k be 5, width be 25ms, and m be 4 in calculation formula (2)), and the signal shown in fig. 2(e) is obtained. The peak of the detection function in the first 10s is obtained to form a sequence, and one half of the median of the sequence is used as the detection function threshold, as shown by the horizontal broken line in the figure.

(6) Presetting a refractory period of 300ms, and regarding a point which simultaneously meets the following three conditions as a starting point of the QRS complex: a. peak points which are larger than the threshold value of the detection function in the detection function obtained in the step (5); b. in the root mean square signal obtained in the step (4), a root mean square peak point which is greater than a root mean square threshold exists in 100ms after the point which meets the condition a; c. the points meeting conditions a and b are outside the preset refractory period. The detection result is shown in fig. 2(f), in which the vertical solid line represents the starting point of the QRS complex of the 12-lead electrocardiographic signal. It can be observed from the figure that the method effectively avoids the starting points of the P wave and the T wave, and accurately detects the average starting point of the QRS wave group of the regular 12-lead sinus rhythm signal.

Example 2: the method for detecting the QRS wave group starting point is applied to the ventricular premature beat signal (before the sinus node impulse reaches the ventricle, the ectopic rhythm point of any part in the ventricle or interventricular interval sends out the electric impulse in advance, so that the depolarization of the ventricle is caused). In this embodiment, a body surface 12 lead electrocardiographic signal with a sampling rate of 1000Hz (i.e. the lead number n is 12) is used, and the working flow is as follows:

(1) the synchronously acquired superficial 12 lead electrocardiographic signals (digital signals) of the polymorphic ventricular premature beat are observed, as shown in fig. 3 (a). The light gray horizontal line of each lead represents the zero potential line, and it can be known from the figure that some leads have serious baseline wander, such as II, III, aVR, aVL, AVF, V1, V3, V4, V5, V6.

(2) And (2) preprocessing the 12-lead electrocardiosignals of the ventricular premature beat in the step (1). Firstly, 10-layer decomposition is carried out on the electrocardiosignals by using sym4 wavelet basis functions, and for the 1000Hz electrocardiosignals, the 10 th-layer decomposition signals (the frequency range is 0-0.98 Hz) are taken as a base line and are set as 0; then, the electrocardiosignals (125-500 Hz) of the 1 st-2 nd layers are subjected to hard threshold processing to remove high-frequency noise. The 12-lead electrocardiographic signal obtained by the above-mentioned preprocessing is shown in fig. 3 (b).

(3) In order to ensure that each lead contributes the same to the final detection result, the preprocessed 12-lead electrocardiosignal obtained in step (2) is normalized to obtain a signal as shown in fig. 3 (c).

(4) And (4) calculating the root mean square of the normalized 12-lead electrocardiosignals obtained in the step (3) to obtain signals shown in a figure 3 (d). The peak value of the root mean square signal in the first 10s is obtained to form a sequence, and one half of the median value of the sequence is taken as the root mean square threshold value, as shown by the horizontal dotted line in the figure.

(5) The detection function of the root mean square signal obtained in step (4) is calculated (let k be 5, width be 25ms, and m be 4 in calculation formula (2)), and the signal shown in fig. 3(e) is obtained. The peak of the detection function in the first 10s is obtained to form a sequence, and one half of the median of the sequence is used as the detection function threshold, as shown by the horizontal broken line in the figure.

(6) Presetting a refractory period of 300ms, and regarding a point which simultaneously meets the following three conditions as a starting point of the QRS complex: a. peak points which are larger than the threshold value of the detection function in the detection function obtained in the step (5); b. in the root mean square signal obtained in the step (4), a root mean square signal peak point greater than a root mean square threshold exists in 100ms after the point meeting the condition a; c. the points meeting conditions a and b are outside the preset refractory period. The detection result is shown in fig. 3(f), in which the vertical solid line represents the starting point of the QRS complex of the 12-lead electrocardiographic signal. It can be observed from the figure that although the ventricular premature heart electrogram has the highly malformed QRS wave, the method still effectively and accurately detects the average starting point of the abnormal QRS complex of the 12-lead electrocardiosignal.

Example 3: the method for detecting the QRS wave group starting point is applied to atrial fibrillation signals (a plurality of rhythms exist, and rhythms at different parts may be inconsistent). In this embodiment, a body surface 12 lead electrocardiographic signal with a sampling rate of 1000Hz (i.e. the lead number n is 12) is used, and the working flow is as follows:

(1) the synchronously acquired body surface 12-lead electrocardiosignals (digital signals) of atrial fibrillation are observed, as shown in fig. 4 (a). The light gray horizontal line of each lead represents the zero potential line, and the electrocardiosignals of some leads are known to have serious baseline drift conditions, such as I, II, aVR, aVL, V3, V5 and V6.

(2) And (3) preprocessing the 12-lead electrocardiosignals of the atrial fibrillation in the step (1). Firstly, performing 10-layer decomposition on signals by using sym4 wavelet basis functions, regarding the electrocardiosignals with 1000Hz, and setting the 10 th-layer decomposition signals (the frequency range is 0-0.98 Hz) as a base line to be 0; then, the electrocardiosignals (125-500 Hz) of the 1 st-2 nd layers are subjected to hard threshold processing to remove high-frequency noise. The 12-lead electrocardiographic signal obtained by the above-mentioned preprocessing is shown in fig. 4 (b).

(3) In order to ensure that each lead contributes the same to the final detection result, the preprocessed 12-lead electrocardiograph signals obtained in step (2) are normalized to obtain signals as shown in fig. 4 (c).

(4) And (4) calculating the root mean square of the normalized 12-lead electrocardiosignals obtained in the step (3) to obtain signals shown in fig. 4 (d). The peak value of the root mean square signal in the first 10s is obtained to form a sequence, and one half of the median value of the sequence is taken as the root mean square threshold value, as shown by the horizontal dotted line in the figure.

(5) The detection function of the root mean square signal obtained in step (4) is calculated (let k be 5, width be 25ms, and m be 4 in calculation formula (2)), and the signal shown in fig. 4(e) is obtained. The peak of the detection function in the first 10s is obtained to form a sequence, and one half of the median of the sequence is used as the detection function threshold, as shown by the horizontal broken line in the figure.

(6) Presetting a refractory period of 300ms, and regarding a point which simultaneously meets the following three conditions as a starting point of the QRS complex: a. peak points which are larger than the threshold value of the detection function in the detection function obtained in the step (5); b. in the root mean square signal obtained in the step (4), a root mean square signal peak point greater than a root mean square threshold exists in 100ms after the point meeting the condition a; c. the points meeting conditions a and b are outside the preset refractory period. The detection result is shown in fig. 4(f), in which the vertical solid line represents the starting point of the QRS complex of the 12-lead electrocardiographic signal. It can be observed from the figure that although the P wave in the electrocardiogram of atrial fibrillation disappears and is replaced by the f wave in a sawtooth shape, the method still effectively and accurately detects the average starting point of the QRS complex of the abnormal 12-lead electrocardiosignal.

Claims

1. A method for detecting the initial point of QRS wave group of multi-lead electrocardiosignal is characterized by comprising the following steps:

(4) calculating the root mean square of the multi-lead normalized electrocardiosignals obtained in the step (3) to obtain RMS signals, and assuming that n-lead electrocardiosignals exist (n is more than or equal to 2), calculating the root mean square as shown in a formula (1):

(5) calculating a detection function DEF using the RMS signal obtained in step (4), the DEF being calculated as shown in equation (2):

setting an o point as a starting point of a QRS wave to be detected, a width as a time parameter and an empirical value as well as k and m as normal numbers and empirical values for the RMS signal of the electrocardiosignal obtained in the step (4); wherein:

c is the maximum value of the RMS signal occurring m width times after point o; it is beneficial to distinguishing QRS wave, P wave and T wave;

d is the time difference between o and the RMS maximum occurring m width times back; the c value is larger but the d value is smaller for ventricular premature beats; the c value of sinus rhythm is small, and the d value is large; therefore, the comprehensive consideration of c and d is beneficial to reducing the difference of the function values of the detection of the non-concentric beats; meanwhile, the d value is also beneficial to distinguishing the QRS wave from the T wave which is partially and slowly increased;

e is the RMS integral value m width time from point o to point o;

(6) calculating the peak value of the root mean square signal in the first seconds by using the data of the first seconds of the root mean square signal obtained in the step (4) to form a group of sequences, and taking the numerical value corresponding to a certain percentage of the median value of the sequences as an RMS threshold value; calculating the peak value of the detection function signal within the first seconds by using the data of the first seconds of the detection function signal obtained in the step (5) to form a group of sequences, and taking the numerical value corresponding to a certain percentage of the median value of the sequences as a DEF threshold value; the DEF peak point satisfying the following three conditions at the same time is taken as the starting point of the QRS complex: a. peak points which are larger than the threshold value of the detection function in the detection function obtained in the step (5); b. in the root mean square signal obtained in the step (4), a root mean square peak point which is greater than a root mean square threshold exists in 100ms after the point which meets the condition a; c. DEF peak points meeting conditions a and b are outside the preset refractory period.